Among members of the model organism community there is almost universal appreciation that the effect of an allele is frequently modulated by genetic background. This concept has been reinforced by the observation that genetically engineered 'knockout mice' usually have different phenotypes when observed on different genetic backgrounds. Although epistasis is clearly one of the potential explanations for the much discussed 'missing heritability,' human geneticists are mostly focusing their attention on rare alleles instead. This is understandable: an unbiased search for interactions in a typical human genome-wide association study (GWAS) would require about 1012 tests. If the effect sizes of the interactions are even smaller than the main effects that have been detected by GWAS, then power to find such interactions will be very poor. Methods for reducing the number of comparisons by prioritizing certain genes may improve power to detect epistatic interactions. We are proposing to take advantage of the highly structured populations that can be created using mouse crosses to detect epistasis and to map it to specific genetic loci. We are focusing our efforts on the gene Cacna1c, which is among the most robust findings to emerge from genome-wide association studies (GWAS) of psychiatric diseases. CACNA1C has been implicated with genome-wide significance for bipolar disorder, and this finding has been replicated in independent samples. More recent data suggest that CACNA1C may also influence the risk for other psychiatric disorders, including major depression and schizophrenia. Mice that are heterozygous for a null allele of Cacna1c show multiple robust behavioral differences. We are proposing to use Cacna1c mutant mice to test our approach. We will cross an inbred C57BL/6J (B6) mouse that is heterozygous for a knockout of Cacna1c with different inbred strains, producing a cohort of F1s in which half of all offspring will inherit one copy of te null allele (+/-) and the other half will inherit two wild-type alleles (+/+). We will then phenotye these mice for behaviors that are known to be altered in Cacna1c +/- mice. Our goal is to determine which F1 backgrounds are susceptible and which ones are resistant to the null allele for each of several behavioral traits. We will use that data to perform a scan for modifiers. Those modifiers can be tested in the human GWAS datasets that were used for the initial identification of CACNA1C. We will use a mixed model to account for the different degrees of relatedness among the inbred strains by using the program QTLRel, which we have developed for the genetic analysis of complex traits. While the proposed studies focus on Cacna1c, our approach is generally applicable to the detection of epistatic interactions with naturally occurring or engineered mutant alleles, provided that they have a dominant mode of inheritance.